Gaussian

Versions Installed

Fionn: 09b01 / 09c01 / 09d01 / 09e01 / 16a03

Description

Starting from the basic laws of quantum mechanics, Gaussian predicts the energies, molecular structures, and vibrational frequencies of molecular systems, along with numerous molecular properties derived from these basic computation types.

License

Gaussian is available for use. Please contact the Helpdesk to gain access.

Benchmarks

N/A.

Thin Component

Gaussian jobs on Fionn are run via shared memory (without Linda) on a single compute node. Users should specify 24 processes in their Gaussian input file (.com or .gjf)

%nproc=24

For good performance use the %mem directive. A good compromise would be 50G. If needed at expense of I/O caches one can go as high as 57G, but expect a few percentages lost in performance compared with the 50G case. Of course you can fine tune these values.

%mem=50000mb

An example of submission PBS script:

#!/bin/bash
#PBS -l nodes=1:ppn=24
#PBS -l walltime=48:00:00
#PBS -N my_job_name
#PBS -A project_name
#PBS -r n
#PBS -j oe
#PBS -m bea
#PBS -M me@my_email.ie

cd $PBS_O_WORKDIR

module load molmodel gaussian/09e01

ulimit -Ss 1048576
ulimit -Sl 524288
ulimit -c 0

g09 < input.gjf > output.log

Shared Memory

Example of script to run on shared memory partition. For each node you select in the -l nodes you get around 120 GiB of RAM for your calculation. Please use %mem inside your gaussian input to specify the amount of memory. Also note that %nproc can be increased in multiples of 8 based on how many nodes you requested.

#!/bin/bash
#PBS -l nodes=3:ppn=8
#PBS -l walltime=72:00:00
#PBS -N my_job_name
#PBS -A project_name
#PBS -r n
#PBS -j oe
#PBS -m bea
#PBS -M me@my_email.ie
#PBS -q ShmemQ

cd $PBS_O_WORKDIR

module load molmodel gaussian/09e01

ulimit -Ss 1048576
ulimit -Sl 524288
ulimit -c 0

g09 < input.gjf > output.log

 

Additional notes

The Gaussian module sets the GAUSS_SCRDIR to the correct location for that node type. Jobs can use very large RWF files as the scratch space is provided by a large high-performance shared volume.

Further information can be obtained at www.gaussian.com.